An ultrasound diagnostic system has been extensively used in the medical field due to its non-invasive and non-destructive nature. The ultrasound diagnostic system is highly safe and produces no dangerous side effects such as exposure to X-rays, etc. However, the ultrasound diagnostic system suffers from inherent shortcomings of an ultrasound image such as a low signal-to-noise ratio and a limited field of view. Thus, the image registration of a CT (or MR) image onto the ultrasound image has been introduced in order to compensate for deficiencies of the ultrasound image.
The image guided intervention is one of the medical applications, which require the image registration between intraoperative and preoperative images. In case of the liver, a 3-dimensional ultrasound image is frequently used for the image-guided intervention. However, due to the poor quality of the 3-dimensional ultrasound image, it is required to display a preoperative 3-dimensional CT (or MR) image, which corresponds to an intraoperative 3-dimensional ultrasound image, but whose quality is relatively high compared thereto. In such a case, the image registration between the 3-dimensional ultrasound image and the 3-dimensional CT image is necessary. The 3-dimensional ultrasound image and the 3-dimensional CT image are formed at different respirations so that local deformation may appear. Thus, a non-rigid image registration, which is an essential step in various medical applications for image registration, should be performed.
However, it is a difficult task to register the 3-dimensional CT image onto the 3-dimensional ultrasound image due to their different imaging characteristics. Several algorithms have been proposed for performing the non-rigid image registration between the 3-dimensional ultrasound and CT (or MR) liver images. For theses image registration, the 3-dimensional ultrasound and CT (or MR) images are converted into vessel probability images. Thereafter, a normalized cross correlation between two vessel probability images is maximized. Also, the image registration method between the 3-dimensional ultrasound and CT (or MR) images, which is performed by extracting and registering centerlines of vessels, has been proposed. In such a method, however, the registration accuracy may depend on vessel segmentation accuracy in the 3-dimensional ultrasound and CT (or MR) images.